deploy.yml

name: Deploy

run-name: >-
  ${{
    github.event_name == 'workflow_run' && format('Deploy dev: {0}', github.event.workflow_run.head_commit.message) ||
    github.event_name == 'release' && format('Deploy prod: {0}', github.event.release.tag_name) ||
    format('Deploy: {0}', github.sha)
  }}

concurrency:
  group: >-
    deploy-${{
      github.event_name == 'release' && format('prod-{0}', github.event.release.tag_name) ||
      github.event_name == 'workflow_run' && format('ci-{0}', github.event.workflow_run.head_branch) ||
      github.ref
    }}
  cancel-in-progress: true

on:
  workflow_run:
    workflows: [CI]
    types: [completed]
    branches: [main]
  release:
    types: [published]
  workflow_dispatch:
    inputs:
      no_cache:
        description: Build without Docker cache
        required: false
        default: false
        type: boolean

env:
  REGISTRY: ghcr.io
  IMAGE_NAME: ${{ github.repository }}

permissions:
  contents: read
  packages: write

jobs:
  deploy:
    name: Deploy
    runs-on: ubuntu-latest
    if: >-
      github.event_name == 'workflow_dispatch' ||
      github.event_name == 'release' ||
      (github.event_name == 'workflow_run' && github.event.workflow_run.conclusion == 'success')

    steps:
      - name: Checkout
        uses: actions/checkout@v4
        with:
          ref: ${{ github.event.workflow_run.head_sha || github.sha }}

      - name: Set deployment vars
        id: vars
        run: |
          if [[ "${{ github.event_name }}" == "release" ]]; then
            echo "image_tag=${{ github.event.release.tag_name }}" >> $GITHUB_OUTPUT
            echo "env_file=.env.prod.enc" >> $GITHUB_OUTPUT
            echo "stack_name=underlay-prod" >> $GITHUB_OUTPUT
          else
            echo "image_tag=${{ github.sha }}" >> $GITHUB_OUTPUT
            echo "env_file=.env.dev.enc" >> $GITHUB_OUTPUT
            echo "stack_name=underlay-dev" >> $GITHUB_OUTPUT
          fi

      - name: Set up Docker Buildx
        uses: docker/setup-buildx-action@v3

      - name: Log in to GHCR
        uses: docker/login-action@v3
        with:
          registry: ${{ env.REGISTRY }}
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - name: Extract metadata
        id: meta
        uses: docker/metadata-action@v5
        with:
          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}

      - name: Build and push
        uses: docker/build-push-action@v6
        with:
          context: .
          target: production
          push: true
          provenance: false
          sbom: false
          no-cache: ${{ inputs.no_cache || false }}
          cache-from: type=gha
          cache-to: type=gha,mode=max
          tags: |
            ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}:${{ steps.vars.outputs.image_tag }}
            ${{ steps.meta.outputs.tags }}
          labels: ${{ steps.meta.outputs.labels }}

      - name: Start SSH agent
        uses: webfactory/ssh-agent@v0.9.0
        with:
          ssh-private-key: ${{ secrets.SSH_PRIVATE_KEY }}

      - name: Install sops
        run: |
          curl -LO https://github.com/getsops/sops/releases/download/v3.9.4/sops-v3.9.4.linux.amd64
          sudo mv sops-v3.9.4.linux.amd64 /usr/local/bin/sops
          sudo chmod +x /usr/local/bin/sops

      - name: Extract DEPLOY_HOST from env file
        id: host
        env:
          SOPS_AGE_KEY: ${{ secrets.SOPS_AGE_SECRET_KEY }}
        run: |
          set -euo pipefail
          DEPLOY_HOST=$(sops -d --input-type dotenv --output-type dotenv "${{ steps.vars.outputs.env_file }}" | grep '^DEPLOY_HOST=' | cut -d= -f2)
          if [[ -z "$DEPLOY_HOST" ]]; then
            echo "::error::DEPLOY_HOST not found in ${{ steps.vars.outputs.env_file }}"
            exit 1
          fi
          echo "deploy_host=${DEPLOY_HOST}" >> $GITHUB_OUTPUT

      - name: Add known hosts
        run: |
          mkdir -p ~/.ssh
          ssh-keyscan -H "${{ steps.host.outputs.deploy_host }}" >> ~/.ssh/known_hosts 2>/dev/null

      - name: Deploy over SSH
        env:
          SSH_USER: ${{ secrets.SSH_USER }}
          DEPLOY_HOST: ${{ steps.host.outputs.deploy_host }}
          REPO: ${{ github.repository }}
          BRANCH: ${{ github.ref_name }}
          GHCR_USER: ${{ secrets.GHCR_USER }}
          GHCR_TOKEN: ${{ secrets.GHCR_TOKEN }}
          IMAGE_TAG: ${{ steps.vars.outputs.image_tag }}
          ENV_FILE: ${{ steps.vars.outputs.env_file }}
          STACK_NAME: ${{ steps.vars.outputs.stack_name }}
        run: |
          ssh "${SSH_USER}@${DEPLOY_HOST}" \
            "env GHCR_USER='${GHCR_USER}' GHCR_TOKEN='${GHCR_TOKEN}' IMAGE_TAG='${IMAGE_TAG}' ENV_FILE='${ENV_FILE}' STACK_NAME='${STACK_NAME}' bash -s -- '${REPO}' '${BRANCH}'" <<'EOS'
          set -euo pipefail

          REPO="${1:?missing repo}"
          BRANCH="${2:-main}"

          : "${IMAGE_TAG:?missing IMAGE_TAG}"
          : "${GHCR_USER:?missing GHCR_USER}"
          : "${GHCR_TOKEN:?missing GHCR_TOKEN}"
          : "${STACK_NAME:?missing STACK_NAME}"

          REPO_NAME="${REPO##*/}"
          APP_DIR="/srv/${STACK_NAME}"
          REPO_SSH="git@github.com:${REPO}.git"

          ssh-keyscan -H github.com >> ~/.ssh/known_hosts 2>/dev/null
          chmod 600 ~/.ssh/known_hosts

          if [[ ! -d "${APP_DIR}/.git" ]]; then
            sudo mkdir -p "${APP_DIR}"
            sudo chown -R "$USER:$USER" "${APP_DIR}"
            git clone --branch "${BRANCH}" "${REPO_SSH}" "${APP_DIR}"
          fi

          cd "${APP_DIR}"
          git fetch --prune --tags origin
          git checkout --detach "${IMAGE_TAG}"

          : "${ENV_FILE:?missing ENV_FILE}"
          umask 077
          sops -d --input-type dotenv --output-type dotenv "$ENV_FILE" > .env

          # Init swarm if not already active
          if ! sudo docker info --format '{{.Swarm.LocalNodeState}}' | grep -qx active; then
            sudo docker swarm init
          fi

          echo "$GHCR_TOKEN" | sudo docker login ghcr.io -u "$GHCR_USER" --password-stdin

          sudo docker pull "ghcr.io/${REPO}:${IMAGE_TAG}"

          # Deploy/update stack — export .env vars for compose interpolation
          # (docker stack deploy does NOT read .env files automatically)
          # IMAGE and IMAGE_TAG are set last to prevent .env from overriding them
          sudo env \
            $(grep -v '^#' .env | grep -v '^$' | xargs) \
            IMAGE="ghcr.io/${REPO}" IMAGE_TAG="${IMAGE_TAG}" \
            docker stack deploy -c docker-compose.yml \
            --with-registry-auth --resolve-image always --prune "${STACK_NAME}"

          # Verify the deployed image matches what was built
          DEPLOYED_IMAGE=$(sudo docker service inspect "${STACK_NAME}_app" --format '{{.Spec.TaskTemplate.ContainerSpec.Image}}')
          echo "Deployed image: ${DEPLOYED_IMAGE}"
          if [[ "$DEPLOYED_IMAGE" != *"${IMAGE_TAG}"* ]]; then
            echo "::warning::Deployed image tag doesn't match expected ${IMAGE_TAG}"
          fi

          sudo docker stack services "${STACK_NAME}"

          # Wait for rollout
          wait_rollout() {
            local svc="$1" timeout="${2:-300}"
            local end=$((SECONDS + timeout))
            while (( SECONDS < end )); do
              local state
              state="$(sudo docker service inspect "$svc" --format '{{if .UpdateStatus}}{{.UpdateStatus.State}}{{end}}' 2>/dev/null || echo "")"
              echo "  $svc: update_state=$state"
              if [[ "$state" == "rollback_started" || "$state" == "rollback_completed" ]]; then
                echo "  ERROR: $svc rolled back!"
                sudo docker service ps "$svc" --no-trunc --format '{{.Name}} {{.CurrentState}} {{.Error}}' | head -10
                return 1
              fi
              if [[ "$state" == "completed" ]] || [[ -z "$state" ]]; then
                # Verify all running tasks are healthy (not just started)
                local unhealthy
                unhealthy="$(sudo docker service ps "$svc" --filter desired-state=running --format '{{.CurrentState}}' 2>/dev/null | grep -cv '^Running' || true)"
                if [[ "$unhealthy" == "0" ]]; then
                  echo "  $svc rollout complete"
                  return 0
                fi
              fi
              sleep 10
            done
            echo "Rollout timeout for $svc"
            sudo docker service ps "$svc" --no-trunc --format '{{.Name}} {{.CurrentState}} {{.Error}}' | head -10
            return 1
          }

          wait_rollout "${STACK_NAME}_app" 300
          wait_rollout "${STACK_NAME}_cron" 120

          # Cleanup old images
          sudo docker image prune -a --filter "until=72h" -f

          echo "Deployed ${STACK_NAME} @ ${IMAGE_TAG} to $(hostname)"
          EOS